Windows 8, 10, 11 and working Stable Diffusion Note: This software is an addition to Stable Diffusion. You need to have any Stable Diffusion running on your system

High Definition Workflow for Stable Diffusion

Why was this created? This was a personal project to create a tool that could fully utilize generative AI in an artist's workflow. The goal was not simply to generate an image and accept it, but rather to be able to modify small details or add objects into a larger composition. However, after finding the tool to be very useful just for my own needs, I decided to make it available for others. What do I need? You need to have a local Stable Diffusion installation and running and be familiar with it. There are various interfaces, the most common being: https://github.com/AUTOMATIC1111/stable-diffusion-webui How do I install Stable Diffusion? If you are asking this question, then probably SDA tool is not for you yet. But don't worry, installation of Stable Diffusion is not difficult as the project offers an easy-to-install solution even for people who are not familiar with Python. Just head to the https://github.com/AUTOMATIC1111/stable- diffusion-webui and read the Installation and Running section. What you need to have is a decent GPU (graphical card). Again, the site will guide you about the minimum requirements. But there are tools in Stable Diffusion to create high resolution images... Of course, there are many scripts and plugins that will automatically upscale the generated image and add details. When you let Stable Diffusion create your image, every detail will be dreamed up and you just need to accept it or try again (re-roll). You can't control all the small details with the prompt alone. This becomes more of a lottery. You are hardly in much control but it's probably fine if you just want to generate an image. SDA Tool was designed for a professional workflow where you need to be in full control over every single part of the image. It's also designed to change just a small part of an existing artwork or composition, which would be quite hard in the SD interface alone, especially when we are talking about large images. No matter what, Stable Diffusion works best at its base resolution of 512 x 512 (or 1024 x 1024 for SD XL) because that was the resolution the images were trained on. SDA uses this limitation and instead of trying to load a large image (let's say 6000x4000) into SD because you need to fix some small detail, it will slice the part and you can work only on the base resolution. Then the part will be seamlessly merged back to the image. Slicing and merging can be done in editing applications like Photoshop... Yes, that's how the project started - from frustration when using Stable Diffusion in conjunction with Photoshop. I would crop a small portion of a larger image, then load it into Stable Diffusion UI to fix or add something, and then place it back into Photoshop, trying to align it with the original and blend it without seams. This is fine if you need to do it once, but working with generative AI is a process where you'd be doing it hundreds of times. There are also some plugins in Photoshop that connect to Stable Diffusion (or use Adobe's own Generative AI if you have a paid subscription), but I found the workflow to be cumbersome, slow, and not particularly enjoyable. SDAssembly was built for my own use because I had tried all the other tools first and was not satisfied. For me this is the fastest workflow. What's Next? At some point, I'm planning to add a simple paint tool to the slices before they are exported – so you can quickly cover some area or paint a crude image which SD can then process into a full detail that fits the scene.